Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for supporting load balancing in a multi-tenant cluster environment, comprising: associating each of a plurality of partitions with one or more nodes of a plurality of end nodes, each of the plurality of end nodes comprising a partition key; providing a plurality of switches, the plurality of switches being arranged in at least two levels; assigning each node of the plurality of nodes a weight parameter; and based at least upon the assigned weight of each of the plurality of nodes, routing the plurality of nodes within the multi-tenant cluster environment, wherein the routing comprises selecting, for each routed node, at least one down-going port and at least one up-going port on at least one switch; wherein the routing preserves partition isolation on ports within the multi-tenant cluster environment.
This invention relates to load balancing in multi-tenant cluster environments, addressing challenges in efficiently distributing workloads while maintaining partition isolation. The method involves associating partitions with multiple end nodes, each containing a partition key, to ensure data isolation. A multi-level switch architecture is used, where switches are arranged in at least two hierarchical levels. Each node is assigned a weight parameter, which influences routing decisions. Based on these weights, nodes are routed within the cluster by selecting down-going and up-going ports on switches. The routing process ensures that partition isolation is preserved across all ports, preventing unauthorized access or interference between different partitions. This approach optimizes load distribution while maintaining security and performance in shared cluster environments. The system dynamically adjusts routing based on node weights, allowing for flexible and scalable load management. The invention is particularly useful in cloud computing and distributed systems where multiple tenants share infrastructure while requiring strict isolation.
2. The method of claim 1 , further comprising: upon selecting the at least one down-going port for the routed node, increasing a load on the down-going port by the assigned weight parameter of the routed node.
This invention relates to network routing, specifically optimizing traffic distribution in a network by dynamically adjusting port loads based on assigned weight parameters. The problem addressed is inefficient traffic routing, where network congestion occurs due to uneven distribution of data across available ports. The solution involves a method for routing data packets through a network by selecting down-going ports for nodes based on their assigned weight parameters, which represent the relative importance or priority of the traffic associated with each node. When a down-going port is selected for a routed node, the load on that port is increased by the node's assigned weight parameter. This ensures that higher-priority traffic receives appropriate bandwidth allocation while maintaining balanced network performance. The method may also include determining the weight parameter for each routed node based on factors such as traffic volume, priority, or network conditions. By dynamically adjusting port loads, the invention prevents congestion and improves overall network efficiency. The system may further include mechanisms to monitor and update weight parameters in real-time to adapt to changing network demands. This approach is particularly useful in large-scale networks where traffic distribution must be carefully managed to avoid bottlenecks and ensure reliable data transmission.
3. The method of claim 2 , further comprising: upon selecting the at least one up-going port for the routed node, increasing a load on the up-going port by the assigned weight parameter of the routed node.
This invention relates to network routing, specifically optimizing traffic distribution in a network by dynamically adjusting port loads based on assigned weight parameters. The problem addressed is inefficient traffic routing, which can lead to congestion, bottlenecks, or suboptimal resource utilization in network nodes. The method involves routing data through a network by selecting at least one up-going port for a routed node, where the selection is based on a weight parameter assigned to the node. This weight parameter influences the distribution of traffic across available ports. When an up-going port is selected, the load on that port is increased by the assigned weight parameter of the routed node. This adjustment ensures that traffic is distributed proportionally according to the weight values, preventing overloading of specific ports while maintaining balanced network performance. The method may also include determining a set of up-going ports for the routed node, where each port in the set has a weight parameter. The selection of a specific port from this set is based on the weight parameter, ensuring that higher-weight nodes contribute more to the load on the selected port. This dynamic adjustment allows for flexible and adaptive routing, improving overall network efficiency and reliability. The approach is particularly useful in scenarios where traffic patterns vary or where certain nodes require prioritized handling.
4. The method of claim 3 , wherein selecting the at least one down-going port for the routed node comprises: comparing a plurality of down-going ports and selecting the down-going port having a least load; in response to two or more ports having the least load, iterating over the two or more ports having the least load and selecting a port having the least load which is part of a switch that is marked with a partition key of the node being routed.
This invention relates to network routing optimization in data center or distributed computing environments. The problem addressed is efficient traffic distribution in multi-path networks, particularly in scenarios where multiple down-going ports (ports leading to lower-tier switches or devices) are available for routing data packets from a routed node. The challenge is to select the optimal down-going port to minimize congestion, balance load, and ensure proper partitioning of network traffic. The method involves selecting a down-going port for a routed node by first comparing the load across multiple available down-going ports. The port with the least load is initially chosen. If multiple ports share the least load, the method further evaluates these ports by checking which of them belongs to a switch marked with the partition key of the node being routed. The port associated with the switch marked with the correct partition key is then selected. This ensures that traffic is not only distributed evenly but also adheres to predefined partitioning rules, which can be critical for maintaining network segmentation, security policies, or performance isolation in distributed systems. The approach optimizes both load balancing and partition-aware routing in multi-path network architectures.
5. The method of claim 4 , wherein assigning each node of the plurality of nodes a weight parameter comprises: receiving the weight parameter from an administrator for each node of the plurality of nodes; and assigning the received weight parameter to each node of the plurality of nodes.
A method for managing node weights in a network or computational system involves dynamically assigning weight parameters to nodes to optimize performance, resource allocation, or decision-making processes. The method addresses the challenge of efficiently distributing influence or priority among interconnected nodes, which is critical in applications such as distributed computing, machine learning, or network routing. The system includes a plurality of nodes, each representing a computational unit, device, or data point, and a controller or processor that manages the assignment of weights to these nodes. The weight parameters determine the relative importance or influence of each node in the system, affecting how data is processed, routed, or prioritized. In this method, the weight parameters are not automatically calculated or derived but are instead manually specified by an administrator. The administrator provides a weight parameter for each node, which is then directly assigned to that node by the system. This manual assignment allows for precise control over node priorities, enabling tailored configurations based on specific operational requirements or constraints. The method ensures that the system adheres to predefined weight distributions, which can be critical in scenarios where administrative oversight is necessary to maintain compliance, security, or performance standards.
6. The method of claim 4 , wherein assigning each node of the plurality of nodes a weight parameter comprises: calculating, via a port data counter, the weight parameter for each node of the plurality of nodes; and assigning the calculated weight parameter to each node of the plurality of nodes.
This invention relates to network traffic management, specifically a method for assigning weight parameters to nodes in a network to optimize traffic distribution. The problem addressed is the inefficient allocation of network resources, leading to congestion and suboptimal performance. The solution involves dynamically calculating and assigning weight parameters to nodes based on traffic data to balance load distribution. The method calculates a weight parameter for each node using a port data counter, which measures traffic volume or other relevant metrics at each node. The calculated weight parameter is then assigned to the corresponding node. This process ensures that nodes with higher traffic loads receive appropriate weight adjustments, improving traffic flow and resource utilization. The weight parameters can be recalculated periodically or in response to traffic changes to maintain optimal performance. The invention may be part of a broader system for network management, where nodes are interconnected and traffic is routed based on these weight parameters. By dynamically adjusting weights, the system avoids bottlenecks and ensures fair distribution of network resources. This approach is particularly useful in high-traffic environments where static configurations fail to adapt to changing conditions. The method enhances network efficiency, reduces latency, and improves overall system reliability.
7. The method of claim 4 , wherein the multi-tenant cluster environment comprises an InfiniBand subnet.
A method for managing network traffic in a multi-tenant cluster environment, particularly one utilizing an InfiniBand subnet, addresses the challenge of efficiently allocating and isolating network resources across multiple tenants while maintaining high performance and security. The method involves dynamically configuring the InfiniBand subnet to support multiple tenants, ensuring that each tenant's network traffic is isolated from others while optimizing resource utilization. This includes assigning dedicated network paths, managing Quality of Service (QoS) parameters, and enforcing security policies to prevent unauthorized access or interference between tenants. The InfiniBand subnet provides high-speed, low-latency communication, making it critical to efficiently partition and allocate its resources to avoid bottlenecks or contention. The method may also involve monitoring traffic patterns, adjusting configurations in real-time, and ensuring compliance with service-level agreements (SLAs) for each tenant. By leveraging the InfiniBand subnet's capabilities, the method ensures that multi-tenant environments can scale efficiently while maintaining performance and security.
8. A system supporting network isolation in a multi-tenant cluster environment, the system comprising: one or more microprocessors; and a processor, running on the one or more microprocessors, wherein the processor operates to perform steps comprising: associating each of a plurality of partitions with one or more nodes of a plurality of end nodes, each of the plurality of end nodes comprising a partition key; providing a plurality of switches, the plurality of switches being arranged in at least two levels; assigning each node of the plurality of nodes a weight parameter; and based at least upon the assigned weight of each of the plurality of nodes, routing the plurality of nodes within the multi-tenant cluster environment, wherein the routing comprises selecting, for each routed node, at least one down-going port and at least one up-going port on at least one switch; wherein the routing attempts to preserve partition isolation on ports within the multi-tenant cluster environment.
The system addresses network isolation challenges in multi-tenant cluster environments, where multiple tenants share infrastructure while requiring secure, isolated communication paths. The system uses a hierarchical switch architecture with at least two levels to manage traffic routing while maintaining partition isolation. Each partition is assigned to one or more end nodes, each containing a partition key to identify its associated tenant. The system assigns a weight parameter to each node, which influences routing decisions. Based on these weights, the system routes nodes by selecting down-going and up-going ports on switches, ensuring that traffic from different partitions remains isolated. The routing algorithm prioritizes preserving partition isolation on ports to prevent unauthorized data leakage between tenants. This approach enhances security and performance in shared cluster environments by dynamically managing network traffic while enforcing strict isolation boundaries. The system is implemented using one or more microprocessors executing the routing logic, ensuring efficient and scalable operation.
9. The system of claim 8 , wherein the processor operates to perform further steps comprising: upon selecting the at least one down-going port for the routed node, increasing a load on the down-going port by the assigned weight parameter of the routed node.
This invention relates to network routing systems, specifically methods for managing traffic load distribution in network nodes. The problem addressed is inefficient load balancing in network routing, where traffic distribution across ports can lead to congestion or underutilization. The system dynamically adjusts traffic flow by assigning weight parameters to nodes and ports to optimize load distribution. The system includes a processor that routes data packets through a network by selecting down-going ports for nodes based on assigned weight parameters. When a down-going port is selected for a routed node, the processor increases the load on that port by the node's assigned weight parameter. This adjustment ensures that the port's load reflects the traffic contribution of the routed node, preventing overloading or imbalance. The system may also track port utilization and dynamically adjust routing decisions to maintain optimal performance. The invention improves network efficiency by dynamically balancing traffic loads across ports, reducing congestion, and ensuring fair resource utilization. It is particularly useful in high-traffic networks where static routing methods fail to adapt to changing conditions. The system's ability to adjust port loads based on node weights allows for more precise traffic management, enhancing overall network performance.
10. The system of claim 9 , wherein the processor operates to perform further steps comprising: upon selecting the at least one up-going port for the routed node, increasing a load on the up-going port by the assigned weight parameter of the routed node.
This invention relates to network routing systems, specifically methods for managing traffic load distribution in network nodes. The problem addressed is inefficient load balancing in network routing, where traffic distribution across ports can lead to congestion or underutilization. The system includes a network node with multiple ports, including at least one up-going port, and a processor that assigns a weight parameter to each routed node (a data packet or flow) to determine its contribution to the load on the selected port. When an up-going port is selected for a routed node, the processor increases the load on that port by the assigned weight parameter of the routed node. This ensures that the load distribution accounts for the relative importance or size of each routed node, improving overall network efficiency and preventing bottlenecks. The system may also include mechanisms for dynamically adjusting weight parameters based on network conditions or traffic patterns to further optimize load balancing. The invention aims to enhance traffic management in network routing by providing a more granular and adaptive approach to load distribution.
11. The system of claim 10 , wherein selecting the at least one down-going port for the routed node comprises: comparing a plurality of down-going ports and selecting the down-going port having a least load; in response to two or more ports having the least load, iterating over the two or more ports having the least load and selecting a port having the least load which is part of a switch that is marked with a partition key of the node being routed.
A system for optimizing network traffic routing in a data center or similar environment addresses the problem of inefficient load distribution across network switches. The system dynamically selects down-going ports for routing data packets to minimize congestion and improve performance. When multiple down-going ports are available for a routed node, the system compares their current load levels and selects the port with the least load. If multiple ports share the same minimum load, the system further evaluates these ports by checking their associated switches. The system prioritizes ports that belong to switches marked with a partition key matching the routed node, ensuring traffic is directed through the most relevant network segments. This approach enhances load balancing and reduces latency by intelligently distributing traffic across the network infrastructure. The system integrates with existing routing protocols to provide adaptive, partition-aware routing decisions.
12. The system of claim 11 , wherein assigning each node of the plurality of nodes a weight parameter comprises: receiving the weight parameter from an administrator for each node of the plurality of nodes; and assigning the received weight parameter to each node of the plurality of nodes.
A system for managing a network of interconnected nodes, where each node is assigned a weight parameter to influence its role or priority in the network. The system allows an administrator to manually set the weight parameter for each node, enabling customization of node behavior based on specific requirements. This weight parameter can be used to prioritize certain nodes over others, adjust resource allocation, or optimize network performance. The system ensures that the administrator-defined weights are accurately applied to each node, providing flexibility in network configuration. This approach is useful in scenarios where nodes have varying importance or capabilities, such as in distributed computing, telecommunications, or IoT networks, where dynamic adjustments are needed to maintain efficiency and reliability. The manual assignment of weights allows for precise control over node behavior, ensuring that the network operates according to predefined administrative policies.
13. The system of claim 11 , wherein assigning each node of the plurality of nodes a weight parameter comprises: calculating, via a port data counter, the weight parameter for each node of the plurality of nodes; and assigning the calculated weight parameter to each node of the plurality of nodes.
This invention relates to a system for managing network traffic in a distributed computing environment, specifically addressing the challenge of efficiently routing data through a network with multiple nodes to optimize performance and resource utilization. The system assigns weight parameters to each node in a network to prioritize traffic flow based on node capacity and performance metrics. The system includes a plurality of nodes interconnected to form a network, where each node is responsible for processing and forwarding data packets. A port data counter is used to monitor traffic at each node, collecting data on packet throughput, latency, and other performance indicators. Based on this data, the system calculates a weight parameter for each node, which reflects its relative importance or capacity in the network. Nodes with higher weight parameters are prioritized for traffic routing, ensuring efficient load balancing and preventing bottlenecks. The weight parameter assignment process involves dynamically adjusting node weights in response to real-time traffic conditions. The port data counter continuously updates performance metrics, allowing the system to recalculate and reassign weights as needed. This adaptive approach ensures that the network remains optimized for varying workloads and traffic patterns. The system may also include additional features, such as failover mechanisms and redundancy protocols, to maintain reliability in case of node failures. Overall, the invention provides a scalable and adaptive solution for managing network traffic in distributed systems.
14. The system of claim 11 , wherein the multi-tenant cluster environment comprises an InfiniBand subnet.
A system for managing a multi-tenant cluster environment includes a network infrastructure that supports multiple tenants with isolated resources. The system ensures secure and efficient communication between tenants by utilizing an InfiniBand subnet, a high-performance networking technology designed for low-latency and high-throughput data transfer. The InfiniBand subnet enables direct memory access (DMA) and remote direct memory access (RDMA) capabilities, allowing tenants to exchange data without involving the CPU, thereby reducing overhead and improving performance. The system further includes mechanisms for resource allocation, monitoring, and security enforcement to prevent unauthorized access between tenants while maintaining high-speed connectivity. The multi-tenant cluster environment is designed to support cloud computing, high-performance computing (HPC), or data center applications where multiple users or organizations share infrastructure while maintaining isolation and performance guarantees. The InfiniBand subnet ensures that tenants can scale their workloads without network bottlenecks, making it suitable for demanding computational tasks such as machine learning, scientific simulations, or large-scale data processing. The system may also include virtualization or containerization techniques to further isolate tenant workloads while leveraging the high-speed networking provided by the InfiniBand subnet.
15. A non-transitory machine readable storage medium having instructions stored thereon for supporting network isolation in a multi-tenant cluster environment that when executed cause a system to perform steps comprising: associating each of a plurality of partitions with one or more nodes of a plurality of end nodes, each of the plurality of end nodes comprising a partition key; providing a plurality of switches, the plurality of switches being arranged in at least two levels; assigning each node of the plurality of nodes a weight parameter; and based at least upon the assigned weight of each of the plurality of nodes, routing the plurality of nodes within the multi-tenant cluster environment, wherein the routing comprises selecting, for each routed node, at least one down-going port and at least one up-going port on at least one switch; wherein the routing preserves partition isolation on ports within the multi-tenant cluster environment.
This invention relates to network isolation in multi-tenant cluster environments, addressing the challenge of securely partitioning network traffic across shared infrastructure. The system uses a non-transitory machine-readable storage medium with instructions to manage network isolation by associating partitions with end nodes, each containing a partition key. A multi-level switch architecture is employed, where switches are arranged in at least two hierarchical levels. Each node is assigned a weight parameter, which influences routing decisions. Based on these weights, nodes are routed within the cluster, with routing involving the selection of down-going and up-going ports on switches. The routing process ensures partition isolation is maintained across all ports, preventing unauthorized traffic between different partitions. The system dynamically assigns and adjusts routing paths to preserve isolation while optimizing network performance. This approach enhances security and efficiency in shared multi-tenant environments by enforcing strict partition boundaries at the network level.
16. The non-transitory machine readable storage medium of claim 15 , the steps further comprising: upon selecting the at least one down-going port for the routed node, increasing a load on the down-going port by the assigned weight parameter of the routed node.
This invention relates to network routing systems, specifically methods for managing traffic load distribution in network nodes. The problem addressed is inefficient load balancing in network routing, where traffic distribution across ports can become unbalanced, leading to congestion and reduced performance. The invention provides a solution by dynamically adjusting the load on down-going ports based on assigned weight parameters for routed nodes. The system involves a non-transitory machine-readable storage medium containing instructions for executing a routing process. When a down-going port is selected for a routed node, the load on that port is increased by the assigned weight parameter of the routed node. This ensures that the port's load reflects the traffic contribution of the routed node, improving load distribution. The weight parameter is a configurable value that represents the relative importance or traffic demand of the routed node, allowing fine-grained control over load balancing. The method also includes determining the routed node's assigned weight parameter, which may be based on factors such as traffic volume, priority, or network policies. The system ensures that the load adjustment is performed in a way that maintains network stability and avoids overloading any single port. This approach enhances network efficiency by dynamically adapting to changing traffic conditions and ensuring balanced utilization of network resources. The invention is particularly useful in large-scale networks where dynamic load balancing is critical for optimal performance.
17. The non-transitory machine readable storage medium of claim 16 , the steps further comprising: upon selecting the at least one up-going port for the routed node, increasing a load on the up-going port by the assigned weight parameter of the routed node.
This invention relates to network routing systems, specifically methods for managing traffic load distribution in network nodes. The problem addressed is the efficient allocation of network traffic across multiple ports to prevent congestion and optimize performance. The invention involves a non-transitory machine-readable storage medium containing instructions for executing a routing process. When a network node is selected for routing, the system assigns a weight parameter to the node, which represents its traffic contribution. The system then selects at least one up-going port (a port directing traffic toward the network core) for the routed node. Upon selection, the system increases the load on the chosen up-going port by the assigned weight parameter of the routed node. This adjustment ensures that the port's load accurately reflects the traffic contribution of the routed node, allowing for balanced traffic distribution and preventing overloading of any single port. The method may also include steps for selecting down-going ports (ports directing traffic away from the network core) and adjusting their loads similarly. The overall system dynamically manages port loads based on node contributions, improving network efficiency and reliability.
18. The non-transitory machine readable storage medium of claim 17 , wherein selecting the at least one down-going port for the routed node comprises: comparing a plurality of down-going ports and selecting the down-going port having a least load; in response to two or more ports having the least load, iterating over the two or more ports having the least load and selecting a port having the least load which is part of a switch that is marked with a partition key of the node being routed.
This invention relates to network routing optimization in data center environments, specifically improving load balancing and partition-aware routing decisions. The problem addressed is inefficient traffic distribution in multi-path networks, where traditional load balancing methods may not account for partition membership, leading to suboptimal routing and potential congestion. The invention describes a method for selecting down-going ports in a network switch during packet routing. When determining the optimal down-going port for a routed node, the system first compares multiple available down-going ports and selects the one with the least current load. If multiple ports share the least load, the system further evaluates these candidates by checking their associated switches. The system prioritizes ports that belong to switches marked with the partition key of the node being routed, ensuring traffic is directed through the most relevant network partition. This approach enhances load balancing while maintaining partition awareness, improving overall network efficiency and reducing congestion. The method integrates with broader routing mechanisms, ensuring that load distribution decisions align with network partitioning strategies. By dynamically assessing both load and partition membership, the system achieves more intelligent routing decisions compared to traditional load-balancing techniques that consider only traffic volume. This solution is particularly valuable in large-scale data center networks where efficient traffic management is critical.
19. The non-transitory machine readable storage medium of claim 18 , wherein assigning each node of the plurality of nodes a weight parameter comprises: receiving the weight parameter from an administrator for each node of the plurality of nodes; and assigning the received weight parameter to each node of the plurality of nodes.
A system and method for managing nodes in a distributed computing environment involves assigning weight parameters to nodes to optimize resource allocation and processing efficiency. The system addresses the challenge of dynamically balancing workloads across nodes with varying capabilities or priorities. Each node in a distributed network is assigned a weight parameter, which determines its influence or priority in task distribution. The weight parameter can be manually set by an administrator for each node, allowing for fine-tuned control over resource allocation. This enables the system to prioritize high-capacity or high-priority nodes, ensuring efficient task distribution and minimizing bottlenecks. The weight parameters are stored in a non-transitory machine-readable storage medium, ensuring persistence and consistency across system operations. By allowing administrators to define node weights, the system provides flexibility in adapting to different workloads, node capabilities, and operational priorities. This approach enhances scalability and performance in distributed computing environments by dynamically adjusting node contributions based on predefined administrative settings.
20. The non-transitory machine readable storage medium of claim 18 , wherein assigning each node of the plurality of nodes a weight parameter comprises: calculating, via a port data counter, the weight parameter for each node of the plurality of nodes; and assigning the calculated weight parameter to each node of the plurality of nodes.
A system and method for optimizing network traffic routing in a distributed computing environment involves dynamically assigning weight parameters to nodes based on their traffic handling capacity. The system addresses the problem of inefficient traffic distribution in large-scale networks, where static routing configurations often lead to congestion and suboptimal performance. The solution calculates a weight parameter for each node using a port data counter, which measures the node's traffic load and processing capability. These weight parameters are then assigned to the nodes to adjust routing decisions, ensuring that traffic is distributed more evenly across the network. The port data counter continuously monitors traffic metrics, allowing the system to dynamically update the weight parameters in real-time. This adaptive approach improves network efficiency by balancing load distribution and reducing bottlenecks. The system is particularly useful in data centers, cloud computing environments, and other large-scale network infrastructures where dynamic traffic management is critical for performance and reliability.
Unknown
June 2, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.